The problem with drawing inferences from one-shot missions, and especially those which return so little engineering telemetry during the critical entry phase, is that you have no idea whether you were brilliant or lucky.

If Beagle 2 had worked, would it have been due to Pillinger's acumen -- i.e., the design would have worked *most* attempts? Or, did we observe a 1 in 100 lucky shot? No way to know for sure with just one, and with no data sent until landing, hard to draw any conclusions whatsoever.

In fact, we still don't know for *sure* with Mars Polar Lander. It's still possible that it functioned flawlessly, but happened to land on the one sharp, pointy boulder in the vincinity.

If we were going to throw a dozen or twenty of a given design at Mars, with the focus quite explicitly on engineering for that first batch, I suspect we could find some places to save money/weight/whatever in the design. I'd bet some parts of the MER system are overdesigned (relative to *actual* needs, not relative to our current uncertain understanding of Mars conditions and system performance) and could be safely squeezed. But there's no easy way to know that without a whole engineering test program.

The problem is that since launches cost so damn much, it's probably cheaper overall to send a handful of overengineered super-expensive landers, than to waste twenty boosters on engineering test articles of cheap landers. If a Delta IV cost just one million bucks, that'd be a different equation.

Of course, I fully agree with Gary's oft-made point that now that we have a system that has passed two tests, let's just send more MERs.